|
|||||||||
|
Thread Tools | Search this Thread |
June 30th, 2010, 08:19 AM | #1 |
Major Player
Join Date: Sep 2008
Location: East Bay Cali
Posts: 563
|
For the Weird Stuff Files Interlace sharpening
I did a weird thing this week and thought that someone else might be interested in it.
when sharpening an intelace video (a lot) many of the alogrythms i tried attempted to cheeply sharpen the whole frame, which results in a bad sharpening. I can try to explain without pictures. You have this fast moving item, and parts of that item are only in one feild, and when a cheap convolution matrix is applied to the whole frame (no dont ask me what that is i dont know :) the "lines" of data that are sticking out the edges of the frame in the one feild ======----- ===----- ====------- just like the above askii pic thing, these loose feild lines against high contrast areas are sharpened, and that is just Wrong, sooo wrong. it increases the interlace artifacts on things that cant Handle a good interlace signal. I had to figure out a way to get the sharpening filters to be "smarter" slip the feild out as a whole item, sharpen that feild, then slap it back together as a frame. With virtual dub i was able to de-interlace using the interlace filter , and do what they call a "side by Side" plopping the fields side by side, instead of interleaved. sharpen that, then slap it back together with the same filter set to return the interlace. I MUST have the interlace, the trick here is not to toss out the motion, but to get HD to SD sharp. in the same filter is the capability to put the side by side back into an interleave of the feilds as it was. by working with the feild as a whole seperated picture (not associated with its frame) it didnt do the sharpening of stuff like the picture above the same , and one set of interlace artifacting didnt exist as bady. you will have to try it someday if you use v-dub, and you will see what i am talking about. world of differance. Untill they get 60 or even 120 Fps going on the video, the interlace is still a good option for fast moving stuff, while many of the digital items just toss out the interlace , having these juttisoned out lines be improperly sharpened due to thier buddy feild the digital things that cant handle interlace wont get as noisey on the fast moving edges. so i can have my cake and eat it also. i can provide a real 60feild video, sharpen it up (say from HD) and it wont get as noisey on the cheap digital junk. Think of it this way , you have a fast moving arm (say), in a feild (for feild sharpening) that arm would be a whole item mostly diaganal across the lines (of the feild), when it is a Frame that line of an arm is broken pieces in one feild of the frame. in the seperated feild squished together as a whole picture the sharpeneing is only along the length of the arm, instead of the broken pieces as seen in the feild. as the motion interlace is tossed out with the bathwater and ignored for its great ability to keep up with about 1/2 of what a human can see, seems the masters of the matrix filtering are making things worse by ignoring its existance. well I dont ($&^@*) when they can do at least 60fps to the customers they can tell us that interlace is now useless. Hey they cant do a 60, so they tell us it isnt needed , what marketing :-) Side By SIde de-interlacing via V-Dubs original filter doubbles the height of the video, which is unnessisary for doing this little trick, so there is a 3rd party filter called VIEWFEILDS and its counterpart UNVIEWFEILDS which will seperate the feilds Top & Bot in similarity, then sharpen and slap it back together and there ya go. some other filters (like resize) will not work without exact even numbers and all, so this is just for convolution (or conveluted) matrixes by putting a lot of sharpening into a HD prior to getting to DV or DVD using this method (instead of the one used by many programs pro and not) you can get a lot nicer psudo HD dvd and still keep the interlace, And have it show better on the junk that cant even properly use interlace. Which lends the question, would interlace be dead if they had not Keeled it :-) but that of course still doesnt solution the issue of my camera (ex1r) not really capable of making a good interlace signal itself due to line summing , soooo that can also be a factor in the whole process. if your still saying to yourself Whaaaaaaatttt the ? well just try it, simply, side by side, sharpen lots, then undo the side by side in V-dub filtering. Then try the same extensive sharpeneing in your $$$$ video program.
__________________
----------------sig----------------- Re-learning everything all over again, one more time. |
June 30th, 2010, 07:58 PM | #2 |
New Boot
Join Date: Jun 2010
Location: USA
Posts: 24
|
I guess what you did was:
de-interlace your clip (separating each field as an independent clip). Apply the filter in question to each independent deinterlaced clip. and then, recombine both re-interlacing. Am I correct? This method (field separation) retain the most quality allowing to apply the filter to each field independently Deinterlacing is needed if you will apply any "filtering" or resampling of pixels (rotations, scaling, blur and sharpening). You need to re-interlace in the same field order as they were de-interlaced. |
July 1st, 2010, 09:28 AM | #3 |
Inner Circle
Join Date: Aug 2005
Location: Atlanta/USA
Posts: 2,515
|
This method has been described in detail right here on this forum by someone (don't have the time to find it right now) as an HD to SD conversion, about 3 years ago.
|
July 2nd, 2010, 09:18 AM | #4 |
Major Player
Join Date: Sep 2008
Location: East Bay Cali
Posts: 563
|
Yes because for some reason 3 of the sharp filters tested decided to apply sharpening (highlites and shadow stuff) to the feild lines juttisoning out into the frame. a terrible thing to do.
__________________
----------------sig----------------- Re-learning everything all over again, one more time. |
| ||||||
|
|